上海交通大学学报

上海交通大学学报 ›› 2026, Vol. 60 ›› Issue (8): 1299-1311.doi: 10.16183/j.cnki.jsjtu.2025.227

• 数字孪生与智能设计 • 上一篇    下一篇

基于表征增强多智能体强化学习的焊接流水车间调度与维护联合优化方法

李洪森a,b,c,d, 张朋b,c,d(), 王明a,b,c,d, 张洁b,c,d, 相文彬b,c,d,e   

  1. a 机械工程学院, 东华大学 上海 201620
    b 纺织工业人工智能技术教育部工程研究中心, 东华大学 上海 201620
    c 人工智能研究院, 东华大学 上海 201620
    d 上海工业大数据与智能系统工程技术研究中心, 东华大学 上海 201620
    e 信息科学与技术学院, 东华大学 上海 201620
  • 收稿日期:2025-07-10 修回日期:2025-10-22 接受日期:2025-11-24 出版日期:2026-08-28 发布日期:2026-09-02
  • 通讯作者: 张 朋,副教授;E-mail:zhangp88@dhu.edu.cn.
  • 作者简介:李洪森(2001—),硕士生,从事机器人调度研究.
  • 基金资助:
    国家自然科学基金(52005099);中央高校基本科研业务费专项(2232025G-14)

Integrated Optimization Method for Scheduling and Maintenance in Welding Flow Shops Based on Representation-Enhanced Multi-Agent Reinforcement Learning

LI Hongsena,b,c,d, ZHANG Pengb,c,d(), WANG Minga,b,c,d, ZHANG Jieb,c,d, XIANG Wenbinb,c,d,e   

  1. a College of Mechanical Engineering, Donghua University, Shanghai 201620, China
    b Engineering Research Center of Artificial Intelligence for Textile Industry of the Ministry of Education, Donghua University, Shanghai 201620, China
    c Institute of Artificial Intelligence, Donghua University, Shanghai 201620, China
    d Shanghai Engineering Research Center of Industrial Big Data and Intelligent System, Donghua University, Shanghai 201620, China
    e College of Information Science and Technology, Donghua University, Shanghai 201620, China
  • Received:2025-07-10 Revised:2025-10-22 Accepted:2025-11-24 Online:2026-08-28 Published:2026-09-02

RichHTML

2

PDF (PC)

458

摘要：

针对带设备预防性维护的焊接流水车间调度问题,以最大完工时间最小化为优化目标,考虑设备故障冲击、有限缓冲区及高维状态空间导致表征模糊等难点,提出基于表征增强多智能体强化学习的调度与维护联合优化方法.将问题拆分为加工调度和预防性维护两个子问题,构建调度-维护双智能体架构;子问题存在强耦合性,其中调度会影响设备故障风险,维护会改变设备可用状态,而双智能体通过价值分解多智能体演员-评论家(VDAC)算法将全局价值函数分解为双智能体的局部价值函数,使两者在优化各自局部价值时自然嵌入对于对方子问题的考量,从而实现协同求解;表征增强通过自编码器提炼高维状态的关键信息,解决了高维状态空间信息冗余、表征模糊问题,使智能体能基于关键表征信息决策,提升调度与维护联合优化性能.算例验证显示,最小化的最大完工时间较其他算法平均减少4.13%,较规则算法平均减少13.34%.

关键词: 流水车间, 多智能体强化学习, 有限缓冲区, 预防性维护, 表征增强

Abstract:

Aiming at the scheduling problem of welding flow shops considering equipment preventive maintenance, with the objective of minimizing the maximum completion time, and addressing challenges such as equipment failure shocks, limited buffers, and ambiguous state representation caused by high-dimensional state spaces, an integrated optimization method for scheduling and maintenance based on representation-enhanced multi-agent reinforcement learning is proposed. The problem is decomposed into two sub-problems: processing scheduling and preventive maintenance, and a scheduling-maintenance dual-agent architecture is constructed. Given the strong coupling between these sub-problems in which scheduling influences equipment failure risks and maintenance alters equipment availability, the dual agents leverage the value-decomposition multi-agent actor-critics (VDAC) algorithm to decompose the global value function into their respective local value functions. This allows both agents to naturally incorporate the influence of the other sub-problems when optimizing their own local objectives, thereby enabling collaborative decision-making. Representation enhancement is introduced via an autoencoder to extract key information from high-dimensional states, alleviating information redundancy and ambiguous representation in high-dimensional state spaces, allowing the agents to make decisions based on key representational information, and improving the performance of joint scheduling and maintenance optimization. Case studies show that, compared with other algorithms, the minimized maximum completion time is reduced by an average of 4.13%, and by an average of 13.34% compared with rule-based algorithms.

Key words: flow shop, multi-agent reinforcement learning, limited buffer, preventive maintenance, representation enhancement

中图分类号: